# AI Engineering Lab: Knowledge Base

The distilled concepts behind the 24-week program. The curriculum points here for
depth; you can also browse it standalone as a reference library.

> **How this knowledge base is built.** Original synthesis by Zorost Intelligence,
> researched from official vendor documentation and primary sources. See
> `research/` for the raw research notes and `14-references-bibliography.md` for
> the full bibliography.

## The concepts

| # | File | Covers |
|---|---|---|
| 01 | [01-ai-engineering-discipline.md](01-ai-engineering-discipline.md) | The AI engineering discipline: Ng's four skills, three loops, roles, and how the program maps to them |
| 02 | [02-ml-dl-fundamentals.md](02-ml-dl-fundamentals.md) | ML & deep learning fundamentals: supervised learning, neural nets, training, evaluation |
| 03 | [03-llm-core-concepts.md](03-llm-core-concepts.md) | Tokens, embeddings, attention, transformers, KV cache, MoE, generation |
| 04 | [04-prompt-context-engineering.md](04-prompt-context-engineering.md) | Prompt engineering and context-window engineering: budgets, compaction, caching, structure |
| 05 | [05-rag-graph-engineering.md](05-rag-graph-engineering.md) | RAG, vector search, chunking, reranking, citations, and knowledge-graph engineering |
| 06 | [06-model-engineering.md](06-model-engineering.md) | Quantization, fine-tuning (SFT/DPO/LoRA), distillation, serving, and selection order |
| 07 | [07-evals-error-analysis.md](07-evals-error-analysis.md) | Evals, LLM-as-judge, traces, human-level parity, error clustering |
| 08 | [08-local-inference-gpu.md](08-local-inference-gpu.md) | Running models locally: Ollama, llama.cpp, MLX, CUDA/MPS/ROCm, VRAM planning |
| 09 | [09-harnesses-tools.md](09-harnesses-tools.md) | Coding-agent harnesses and routers: Claude Code, Cursor, OpenCode, DeepSeek Harness, OpenRouter |
| 10 | [10-agents-multiagent.md](10-agents-multiagent.md) | Agent fundamentals, multi-agent orchestration, memory, guardrails |
| 11 | [11-mcp-ecosystem.md](11-mcp-ecosystem.md) | MCP, A2A, and the agent connectivity ecosystem |
| 12 | [12-cloud-platforms.md](12-cloud-platforms.md) | Azure AI Foundry, Google Vertex AI, AWS Bedrock: capabilities and comparison |
| 13 | [13-databricks-overview.md](13-databricks-overview.md) | The Databricks Data Intelligence Platform at a glance |
| 14 | [14-references-bibliography.md](14-references-bibliography.md) | Textbooks, papers, letters, and vendor docs the program draws from |

## Research notes (raw)

| File | Topic |
|---|---|
| [research/00-zorost-brand-notes.md](research/00-zorost-brand-notes.md) | Zorost Intelligence branding collected from zorost.com |
| [research/01-source-repo-analysis.md](research/01-source-repo-analysis.md) | Analysis of a representative open LLM-engineering course repo |
| [research/02-ng-framework-textbooks.md](research/02-ng-framework-textbooks.md) | Ng's skills map & letters, DeepLearning.AI catalog, textbooks |
| [research/03-harnesses-local-stacks.md](research/03-harnesses-local-stacks.md) | Harnesses, routers, local inference, quantization, fine-tuning stacks |
| [research/04-agents-ecosystem.md](research/04-agents-ecosystem.md) | Agents, multi-agent, MCP/A2A, OpenClaw, Hermes, frameworks |
| [research/05-cloud-platforms.md](research/05-cloud-platforms.md) | Azure AI Foundry, Google Vertex AI, AWS Bedrock deep research |
| [research/06-databricks-deep-dive.md](research/06-databricks-deep-dive.md) | Databricks zero-to-hero research (the Databricks module backbone) |
| [research/07-zorost-skills-map-guide.md](research/07-zorost-skills-map-guide.md) | Zorost's own published analysis of Ng's skills map (six-block programme, systems spine) |
| [research/08-zorost-fieldwork-notes.md](research/08-zorost-fieldwork-notes.md) | Zorost AI Fieldwork: methodology, experiments, reporting standard |

---
© 2026 Zorost Intelligence LLC · https://zorost.com